IRC channel logs


back to list of logs

<civodul>rekado: thanks a lot for going this far
<civodul>apologies for not going anywhere :-/
<rekado>python-jupyterlab-server and python-doubletdetection both fail with an error like this: ERROR: jupyterlab-server==2.12.0 InvalidRequirement('Expected end or semicolon (after version specifier)\n tinycss2>=1.1.0<1.2\n ~~~~~~~^')
<rekado>I don’t see any mention of tinycss2 in the code, so this must be generated from the inputs
<rekado>I suppose one of the inputs has an invalid requirements list
<rekado>python-ipydatawidgets fails in the same way
<rekado>they all have python-nbconvert in common
<zimoun>And again, « I neeed to install Conda because X »
<zimoun>where X is already part of Guix.
<zimoun>But only because someone™ said « Install Conda and done »
<zimoun>Argh, and I have spent hours to explain how to use Guix.
<zimoun>I am sad. )-:
<zimoun>Wow! 861M
<zimoun>tiny script that I can audit ;-)
<zimoun>Next line…
<zimoun>517 lines of Bash. Ahah!
<rekado>this is a common way of distributing self-installing tarballs
<rekado>there’s a preamble that’s the actual script, a separator, and then binary data (often a tarball)
<rekado>the script cuts itself in two and unpacks the tarball
<zimoun>yeah, more than opaque. ;-)
<zimoun>rekado: Recently, you were doubtful about the usefulness of your work around bootstrapping GHC. Here a blog post that shows you opened the road. :-)
<rekado>I tried something different back then: split modules so that they are no longer mutually recursive
<rekado>this got me past the initial hurdle, but I ran into a different wall: GHC depends on types implemented in the GHC runtime system (RTS), and Hugs cannot use GHC’s RTS.
<efraim>i saw that blog post too!
<rekado>(that idea was sent to me by joey hess)
<rekado>my notes mention merging modules to avoid recursive module dependencies:
<zimoun>Joey Hess, everywhere. :-)
<zimoun>rekado: thanks for your notes.
<zimoun>Well, if I read correctly, 6.12.3 and 7.4.2 are missing to have a complete chain from 4.08.2 to 9.2.5 (or 9.4.4).
<zimoun>however 4.08.2 built using old GCC still needs outputs from previous GHC.
<rekado>I haven’t been able to build more recent GHCs with the existing chain from 4.08.2
<zimoun>Do you mean build 6.12.3 using 6.10.4?
<rekado>whichever is the next one; I forget
<rekado>I just looked at my WIP stash for haskell: yes, I failed to build 6.12.3 with ghc-6.10
<zimoun>Damned! Why there are so many blocks on the road? :-)
<civodul>you'd hope that GHC developers would think about a bootstrapping technique that does not involve "archeology"
<civodul>at least so they can compete with OCaml
<rekado>I don’t think they have any interest in anything other than pushing forward.
<rekado>ACTION fixes pytorch
<rekado>oh, bioc 3.17 is out